Académique Documents
Professionnel Documents
Culture Documents
SCIENCE,
BIG
SCIENCE
DEREK J. DE SOLLA PRICE
COLUMBIA
UNIVERSITY PRESS
PREFACE
I gratefully acknowledge the stimulus and help aflForded by
graduate students in my seminars at Yale University, the
1. PROLOGUE TO A SC IE N C E OF SC IE N C E
work and devotion of my research assistant, Miss Joy Day,
and the intent hterary vigilance of my good friend David
2. G A LTO N R E V ISIT E D
33
Klein. Thanks are also due to Asger Aaboe for drawing many
beautiful graphs. I am further indebted to Yale University 3. IN V ISIB L E COLLEGES AND TH E AFFLUENT
Press for permission to reproduce figures from Science Since SC IE N T IFIC COM M UTER 62
Babylon which I published with them in 1962; to the McGraw-
Hill Book Company for permission to reproduce the graph on 4. P O L IT IC A L STRA TEGY FOR BIG SC IE N T IST S 92
development of accelerators from M. S. Livingston and J. P.
Blewett, Particle Accelerators; to the Addison-Wesley Publish IN D EX 116
ing Company for the graph on growth of cities from G. K. Zipf,
Human Behavior and the Principle of Least E§ort; and to the
Cambridge University Press for permission to use an adaption
of a figure from D’Arcy W. Thompson, Growth and Form.
( 1)
How big are you, baby? PROLOGUE TO
Why, don’t you know. A SCIENCE
You’re only so big. OF S C I E N C E
And there’s still room to grow.
( n u r se r y rh ym e)
DURING A M EETING at which a number of great physicists were
to give firsthand accounts of their epocli-making discoveries,
the chairman opened the proceedings with the remark: “Today
we are privileged to sit side-by-side with the giants on whose
shoulders we stand ” ^ This, in a nutshell, exemplifies the pe
culiar immediacy of science, the recognition that so large a
proportion of everything scientific that has ever occurred
is happening now, within living memory. To put it another
way, using any reasonable definition of a scientist, we can
say that 80 to 90 percent of all the scientists that have ever
lived are alive now. Alternatively, any young scientist, starting
^ Gerald Holton, “On the recent past of physics,” American Journal of
Physics, 29 (December, 1961), 805. I should like to draw attention to
the fine study published while this work was in progress: Gerald Holton,
“ Models for Understanding the Growth and Excellence of Scientific
Research,” in S. R. Graubard and G. Holton, eds.. Excellence and
Leadership in a Democracy (New York, Columbia University Press,
1962), 94-131, first published as “Scientific research and scholarship:
notes towards the design of proper scales,” in Proceedings of the Ameri
can Academy of Arts and Sciences, 91 (No. 2 ), 362-99 (Daedalus,
March, 1962). My work derives much from this previous publication,
though its author and 1 do not always agree in detail in the conclusions
we derive from the statistical data.
2 A SC IE N C E OF SC IE N C E A SC IE N C E OF SC IE N C E 3
now and looking back at the end of his career upon a normal Big Science is so diflFerent from the former state of affairs
life span, will find that 80 to 90 percent of all scientific work that we can look back, perhaps nostalgically, at the Little
achieved by the end of the period will have taken place before Science that was once our way of life.
his very eyes, and that only 10 to 20 percent will antedate his If we are to understand how to live and work in the age
experience. newly dawned, it is clearly necessary to appreciate the nature
So strong and dominant a characteristic of science is this of the transition from Little Science to Big Science. It is only
immediacy, that one finds it at the root of many attitudes too easy to dramatize the change and see the differences
taken by scientist and layman toward modern science. It is with reckless naiVet^. But how much tmth is there in the
what makes science seem essentially modem and contempora picture of the Little Scientist as the lone, long-haired genius,
neous. As a historian of science, I find myself doing annual moldering in an attic or basement workshop, despised by so
battle to justify and uphold the practice of spending more ciety as a nonconformist, existing in a state of near poverty,
than half our time on the period before Newton, whereas motivated by the flame burning within him? And what about
every contemporary scientist around knows that what really the corresponding image of the Big Scientist? Is he honored
counts is science since Einstein. in Washington, sought after by all the research corporations
Because the science we have now so vastly exceeds all of the “Boston ring road,” part of an elite intellectual brother
that has gone before, we have obviously entered a new age hood of co-workers, arbiters of political as well as technological
that has been swept clear of all but the basic traditions of destiny? And the basis of the change—was it an urgent public
the old. Not only are the manifestations of modem scientific reaction to the first atomic explosion and the first national
hardware so monumental that they have been usefully com shocks of military missiles and satellites? Did it all happen
pared with the pyramids of Egypt and the great cathedrals of very quickly, with historical roots no deeper in time than the
medieval Europe, but the national expenditures of manpower Manhattan Project, Cape Canaveral rocketry, the discovery of
and money on it have suddenly made science a major segment penicillin, and the invention of radar and electronic computers?
of our national economy. The large-scale character of modem I think one can give a flat “No” in answer to all these
science, new and shining and all-powerful, is so apparent that questions. The images are too naively conceived, and the transi
the happy term “Big Science” has been coined to describe tion from Little Science to Big Science was less dramatic and
it.^ Big Science is so new that many of us can remember its more gradual than appears at first. For one thing, it is clear
beginnings. Big Science is so large that many of us begin to that Little Science contained many elements of the grandiose.
worry about the sheer mass of the monster we have created. And, tucked away in some academic comers, modem Big
* Alvin M. Weinberg, “Impact of large-scale science on the United Science probably contains shoestring operations by unknown
States,” Science, 134 (July 21, 1961), 164. I am indebted to this paper pioneers who are starting lines of research that will be of
for many ideas. See also further comments by Weinberg in “The Federal decisive interest by 1975. It is the brave exception rather than
Laboratories and science education,” Science, 136 (April 6, 1962), 27.
4 A SC IE N C E OF SC IE N C E A SC IE N C E OF SC IE N C E &
the rule that key break-throughs are heralded at birth as im mode of growth is exponential. That is to say, science grows
portant work done by important people. at compound interest, multiplying by some fixed amount in
Historically, there have been numerous big national eflPorts; equal periods of time. Mathematically, the law of exponential
the great observatories of Ulugh Beg in Samarkand in the growth follows from the simple condition that at any time
fifteenth century, of Tycho Brahe on his island of Hven in the the rate of growth is proportional to the size of the population
sixteenth century, and of Jai Singh in India in the seventeenth or to the total magnitude already achieved—the bigger a
century, each of which absorbed sensibly large fractions of thing is, the faster it grows. In this respect it agrees with the
the available resources of their nations. As international efforts, common natural law of growth governing the number of
there were the gigantic expeditions of the eighteenth century human beings in the population of the world or of a particular
to observe the transits of Venus. And, as large-scale hardware, country, the number of fruit flies growing in a colony in a
there were the huge electrical machines, produced most notably bottle, or the number of miles of railroad built in the early
in Holland in the eighteenth century, machines that in their industrial revolution.
time seemed to stretch man’s scientific engineering to its ulti It might at first seem as if establishing such an empirical
mate capability and to give him the power to manufacture the law of growth for science was neither unexpected nor signifi
most extreme physical forces of the universe, rivaling the very cant. The law has, however, several remarkable features, and
lightning and perhaps providing keys to the nature of matter from it a number of powerful conclusions can be drawn. In
and of life itself. In a way, our dreams for modem accelerators deed, it is so far-reaching that I have no hesitation in sug
pale by comparison. gesting it as the fundamental law of any analysis of science.
But let us not be distracted by history. What shall concern Its most surprising and significant feature is that, unlike
us is not so much the offering of counterexamples to show most pieces of curve-fitting, the empirical law holds true with
that Little Science was sometimes big, and Big Science little, high accuracy over long periods of time. Even with a some
but rather a demonstration that such change as has occurred what careless and uncritical choice of the index taken as a
has been remarkably gradual. To get at this we must begin measme, one has little trouble in showing that general ex
our analysis of science by taking measurements, and in this ponential growth has been maintained for two or three cen
case it is even more diflScult than usual to make such deter turies. The law therefore, though at this stage still merely
minations and find out what they mean. empirical, has a status immediately more significant than the
Our starting point will be the empirical statistical evidence usual short-term economic time series. This leads one to a
drawn from many numerical indicators of the various fields and strong suspicion that the law is more than empirical—and
aspects of science. All of these show with impressive con that with suitable definitions of the indices that grow expo
sistency and regularity that if any suflSciently large segment nentially, one may show, as we later shall, that there is a reason
of science is measured in any reasonable way, the normal able theoretical basis for such a law.
KJ A b Ul KN C E OF S C I E NC E A SC IE N C E OF SC IE N C E /
A second important feature of the growth of science is that Important physicists
it is surprisingly rapid however it is measured. An exponential Number of chemical elements known
Accuracy of instruments
increase is best characterized by stating the time required College entrants/1000 population
for a doubling in size or for a tenfold increase.® Now, de
pending on what one measures and how, the crude size of 15 years
science in manpower or in publications tends to double within B.A., B.Sc.
a period of 10 to 15 years. The 10-year period emerges from Scientific journals
Membership of scientific institutes
those catchall measures that do not distin^uisbJAw-gradfi..wQrk Number of chemical compounds known
from high but adopt a basic, minimal definiti^ the Number of scientific abstracts, all fields
15-year period results when one is more selective, counting
only some more stringent defijnipjgn^nf published snmntifin work 10 years
and those who produce it. If this stringency is increased so Number of asteroids known
Literature in theory of determinants
that only scientific work of uery high gnghty^ is counted, then Literature in non-Euclidean geometry
the doubhng period is drawn out so that it approaches about Literature in x rays
20^ a r s . Literature in experimental psychology
The following list shows the order of magnitudes of an as Number of telephones in United States
Number of engineers in United States
sortment of measurable and estimatable doubling times and Speed of transportation
shows how rapidly the growth of science and technology has Kilowatt-hours of electricity
been outstripping that of the size of the population and of
our nonscientific institutions. 5 years
Number of overseas telephone calls
100 years Magnetic permeability of iron
Entries in dictionaries of national biography 1 Y2 years
50 years Million electron volts of accelerators
L abor force
Population Bearing in mind the long period of validity of exponential
Num ber of universities growth, let us note that a 15-year doubling time extended over
20 years three centuries of growth corresponds to an increase of 20
Gross National Product powers of two, or a factor of about one million. Thus, in the
Important discoveries interval from 1660 to the present day, such indices of the size
®It is easy enough to convert from one to the other by noting, as a of science should have increased by the order of a million. To
rough approximation, that 10 doubling periods correspond to a factor
of 1024, or about 3 tenfolding periods. offer the soundest explanation of the scientific and industrial
8 A SC IE N C E OF SC IE N C E
revolutions is to posit that this is indeed what has been hap
pening.
Just after 1660, the first national scientific societies in the
modem tradition were founded; they established the first
scientific periodicals, and scientists found themselves begin
ning to write scientific papers instead of the books that hitherto
had been their only outlets. We have now a world list of some
50.000 scientific periodicals (Fig. 1) that have been founded,
of which about 30,000 are still being published; these have
produced a world total of about six million scientific papers
(Fig. 2) and an increase at the approximate rate of at least
half a million a year.^ In general, the same applies to scientific
manpower. Whereas in the mid-seventeenth century there were
a few scientific men—a denumerable few who were countable
and namable—there is now in the United States alone a popula
tion on the order of a million with scientific and technical de
grees (Fig. 3). What is more, the same exponential law ac
counts quite well for all the time in between. The present
million came through intermediate stages of 100,000 in 1900,
10.000 in 1850, and 1000 in 1800. In terms of magnitude
alone, the transition from Little Science to Big Science has
been steady—or at least has had only minor periodic fluctua Fig. 1. TOTAL NUM BER OF SCIEN TIFIC JO URN ALS AND ABSTRACT
tions similar to those of the stock market—and it has fol JO U RN ALS FOUNDED, AS A FUNCTION O F DATE
lowed a law of exponential growth with the time rates pre
Note that abstracts begin when the population of journals is approxi
viously stated. mately 300. Numbers recorded here are for journals founded, rather than
Thus, the steady doubling every 15 years or so that has those surviving; for all periodicals containing any “science” rather than
for “strictly scientific” journals. Tighter definitions might reduce the
brought us into the present scientific age has produced the absolute numbers by an order of magnitude, but the general trend re
peculiar immediacy that enables us to say that so much of mains constant for all definitions. From Derek J. de Sofia Price, Science
Since Babylon (New Haven, Yale University Press, 1961).
*F or a more detailed discussion of this see Derek J. de Sofia Price,
Science Since Babylon (New Haven, Yale University Press, 1961), Chap
ter 5.
A SC IE N C E OF SC IE N C E li
science is current and that so many of its practitioners are
alive. If we start with the law that the number of living scien
tists doubles in, let us say, 15 years, then in any interval of
15 years there will come into being as many scientists again
as in the whole of time preceding. But at any moment there
coexists a body of scientists produced not over 15 years but
over an interval nearer to the 45 years separating average
date of arrival at the research front from average date of re
tirement from active scientific work. Thus, for every one per
son bom before such a period of 45 years, we now have one
bom in the first doubling period, two in the second, and four
in the third. There are, then, about seven scientists alive for
every eight that have ever been, a fraction of 8 7 ^ percent;
let us call this a coefficient of immediacy. One may calculate
this exactly by using actuarial mortality tables, but in fact
the result is not much altered by this because the doubling
period of science is so much less than the average working
life of a scientist.
For a doubling period of 10 years, the corresponding co
efficient of immediacy is about 96 percent; for a doubling time
of 20 years, about 81 percent. Thus, even if one admits only
the general form of the growth function and the order of
magnitude of its time constant, these account for the feeling
that most of the great scientists are stiU with us, and that the
greater part of scientific work has been produced within living
memory, within the span of the present generation of scientists.
Furthermore, one can emphasize the principle by remarking
that some time between the next decade and the one after we
It will be noted that after an initial period of rapid expansion to a stable shall have produced as much scientific work and as many
growth rate, the number of abstracts increases exponentially, doubling
in approximately 15 years. scientists as in the whole of time up to the present.
What we have said so far is by now well known and reason-
A SC IE N C E OF SC IE N C E 13
ably well agreed upon by those who speculate about science
for fun or high policy. I should like to extend these results,
however, in a couple of ways that may suggest that this out
look requires revision. In the first place, speaking in terms of
a "coefiicient of immediacy” can be misleading. Let us com
pare the figures just found with the conjectural figures for
world population.
At the beginning of the Christian Era, the human race
numbered about 250 million; it grew slowly and erratically,
differently in different places and at different times, and
reached a figure of 550 million by the mid-seventeenth cen
tury. It has grown at an ever-increasing pace, so that today
there are about 3000 million people, and it looks as though
that number will double every 40 to 50 years. If we reckon
about 20 years to a generation, there must have been at least
60,000 millions of people, and thus only about 5 percent of
those who have lived since the beginning of our era are alive
now. If we count all those who lived before the time of Christ,
the fraction will be smaller; if we count only those who have
lived since the mid-seventeenth century, it will be a little
more than 10 percent. Making due allowance for changing
mortalities and age of child-bearing will not, I feel, materially
alter the qualitative result that the human population is far
from immediate in the sense that science is.
Even if we accept the gloomy prognostications of those
who talk about the admittedly serious problem of the popula
POPULATION IN TH E UNITED STATES tion explosion, it would apparently take about another half-
It may be seen that the more highly qualified the manpower, the greater century—some time after the year 2000—^before we could
has been its growth rate. It will also be noted that there appears a claim that 50 percent of all the human beings that have lived
distinct tendency for the curves to turn toward a ceiling value running
parallel with the population curve. were at that moment alive. Most of the persons that have
ever lived are dead, and, in the sense that this will continue
14 A SC IE N C E OF SC IE N C E A SC IE N C E OF SC IE N C E 15
to be so, they will stay dead. One might conclude, since the the seventeenth century. Science has always been modern; it
rate of growth of entries in the great dictionaries of national has always been exploding into the population, always on the
biographies shows a fairly constant proportion to the popula brink of its expansive revolution. Scientists have always felt
tion at various dates, that most of the great or worthy persons themselves to be awash in a sea of scientific literature that
of the world are dead. That is why history is a subject rather augments in each decade as much as in all times before.
different from history of science. There is much more past It is not difficult to find good historical authority for this
to live in if you discuss politics and wars than if you discuss feeling in all epochs. In the nineteenth century we have
science. Charles Babbage in England and Nathaniel Bowditch in the
The immediacy of science needs a comparison of this sort United States bitterly deploring the lack of recognition of the
before one can realize that it implies an explosion of science new scientific era that had just burst upon them. In the eight
dwarfing that of the population, and indeed all other explo eenth century there were the first furtive moves toward special
sions of nonscientific human growth. Roughly speaking, every journals and abstracts in a vain attempt to halt or at least
doubling of the population has produced at least three dou rationalize the rising tide of publications; there is Sir Hum
blings of the number of scientists, so that the size of science phrey Davy, whose habit it was to throw books away after
is eight times what it was and the number of scientists per reading on the principle that no man could ever have the
million population has multiplied by four. Mankind’s per capita time or occasion to read the same thing twice. Even in the
involvement with science has thus been growing much more seventeenth century, we must not forget that the motivating
rapidly than the population. purpose of the Philosophical Transactions of the Royal Society
A second clarification, one of crucial importance, must be and the Journal des Sgavans was not the publishing of new
made concerning the immediacy and growth of modem science. scientific papers so much as the monitoring and digesting of
We have already shown that the 80- to 90-percent currency of the learned publications and letters that now were too much
modem science is a direct result of an exponential growth that for one man to cope with in his daily reading and corre
has been steady and consistent for a long time. It follows that spondence.^
this result, true now, must also have been true at all times The principle of more than 80 percent being contempora
in the past, back to the eighteenth century and perhaps even neous is clearly sufficient to cast out any naive idea that sheer
as far back as the late seventeenth. In 1900, in 1800, and per change in scale has led us from Little Science to Big Science.
haps in 1700, one could look back and say that most of the If we are to distinguish the present phase as something new,
scientists that have ever been are ahve now, and most of what something different from the perception of a burgeoning science
is known has been determined within hving memory. In that
respect, surprised though we may be to find it so, the scientific ®An excellent historical account of the birth of scientific journals is
given by David A. Kronick, A History of Scientific and Technical Periodi
world is no different now from what it has always been since
cals (New York, Scarecrow Press, Inc., 1962).
16 A SC IE N C E OF SC IE N C E A SC IE N C E OF SC IE N C E 17
that was common to Maxwell, to Franklin, and to Newton, left in a quandary. To escape from it one may be tempted at
then we cannot rest our case on the rate of growth alone. A first to deny that there has been any such radical transforma
science that has advanced steadily through more than five tion of the state of science. This is amply belied by the fact
orders of magnitude in more than 250 years is not going to be that since World War II we have been worried about ques
upset by a mere additional single order of magnitude such tions of scientific manpower and literature, government spend
as we have experienced within the last few decades of the ing, and military power in ways that seem quite different, not
present century. merely in scale, from all that went before.
As a side point one may note that the constancy of this Even if one admits that new things are happening and that
phenomenon of immediacy is typical of many other constancies Big Science differs not merely in scale from Little Science,
in science that make it meaningful and useful to pursue the one might still maintain that it was the cataclysmic changes
history of science even though most of our past is alive. What associated with World War II that initiated us into the new
we must do in the humanistic and the scientific analyses of era and produced all the major changes. Quite unexpectedly,
science is search out such constancies of scientific method, of one can show from the statistical studies we have been using
public reaction, of the use of matheniatical models or euphoric to measure the pure growth that the influence of the war on
hardware or the groimd rules of manpower and motivation, scientific manpower and literature seems only to have been
and apply them to our criticism and understanding of this the production of a temporary perturbation that extended for
science that seems so essentially modem and out of all rela its duration.
tion to Archimedes or Galileo or Boyle or Benjamin Franklin. For this interval it is not possible to use the indices one
If we honor a Boyle for his law, or a Planck for his constant, might use before or after; manpower may be in military service,
this is largely accidental hero worship; more important to publication may be suppressed for secrecy. Yet it is apparent
us than the names of those who have quarried a slab of im that the exponential increase after the war is identical with
mortality is their having done so in a manner which notably that before (Fig. 4 ). This is a strong result, for it shows that
illustrates the constant and seemingly eternal way in which the percentage increase per annum is the same before and
these things have been going on. To take an early example after the war and, therefore, if there is any constancy about
such as GaHleo, seen in all its historical perspective, is in the way in which scientific papers generate new scientific
many ways more eflBcient than choosing a recent example such papers and researchers generate new classes of researchers,
as Oppenheimer, though Galileo can tell us nothing of the there cannot have been any great loss or gain to science during
content of modem atomic physics as can Oppenheimer. the war. With the exception only of a sidewise displacement
To return to our main point, if the sheer growth of science of the curve due to secrecy loss, science is just where it would
in its exponential climb is not admissible as an explanation have been, statistically speaking, and is growing at the same
for the transition from Little Science to Big Science, we are rate as if there had been no war. The order of events might
1 8 A S C IE N C E O F S C IE N C E A SC IE N C E OF SC IE N C E 19
have been dijBFerent, the pohtical implications perhaps grossly war looms as a huge milepost, but it stands at the side of a
so, but there is some reason for taking a fatalistic line that straight road of exponential growth.
it was in the nature of things for accelerator laboratories to If, then, we are to analyze the peculiarities of Big Science,
grow as large as Brookhaven, and missile establishments as we must search for whatever there is other than the steady
large as Cape Canaveral, and that had there been no Man hand-in-hand climb of all the indices of science through suc
hattan Project there might still have been a Sputnik. The cessive orders of magnitude. There are, I propose, two quite
different types of general statistical phenomena of science-in-
Thousands of “Physics Abstracts" since 1900 the-large. On the one hand, although we have the over-all
picture of a steady exponential growth with this amazingly
short time constant of about 15 years, not all things are grow
ing at precisely this rate; some are faster, others slower, though
all of them outpace the growth of the population. On the
other hand, we have the possibility that the exponential law
of growth may be beginning to break down.
It is just possible that the tradition of more than 250 years
represents a sort of adolescent stage during which every half-
century science grew out of its order of magnitude, donned
a new suit of clothes, and was ready to expand again. Per
haps now a post-adolescent quiescence has set in, and such
exuberant growth has slowed down and is about to stop upon
the attainment of adult stature. After all, five orders of magni
tude is rather a lot. Scientists and engineers are now a couple
of percent of the labor force of the United States, and the
annual expenditure on research and development is about
the same fraction of the Gross National Product. It is clear
that we cannot go up another two orders of magnitude as we
have climbed the last five. If we did, we should have two
Fig. 4. TOTAL NUM BER OF PHYSICS ABSTRACTS PUBLISHED SINCE scientists for every man, woman, child, and dog in the popula
JANUARY 1, 1900 tion, and we should spend on them twice as much money as
The full curve gives the total, and the broken curve represents the ex
we had. Scientific doomsday is therefore less than a century
ponential approximation. Parallel curves are drawn to enable the effect distant.
of the wars to be illustrated. From Derek J. de SoUa Price, Science Since At a later point I shall treat separately the problem of
Babylon (New Haven, Yale University Press, 1961).
20 A SC IE N C E OF SC IE N C E A SC IE N C E OF SC IE N C E 21
growths at rates diflFerent from that of basic exponential in not go in its accustomed fashion (Fig. 5). In its typical pattern,
crease. We shall consider such growths as slowly changing growth starts exponentially and maintains this pace to a point
statistical distributions of the indices rather than as separate almost halfway between floor and ceiling, where it has an in
rates of increase. Thus, for example, if the number of science flection. After this, the pace of growth declines so that the curve
Ph.D’s were doubling every 15 years, and the number of good continues toward the ceiling in a manner symmetrical with the
ones only every 20 years, the quota of Ph.D’s per good physicist way in which it climbed from the floor to the midpoint. This
would be doubling only every 60 years, a change so slow that symmetry is an interesting property; rarely in nature does one
we can count it out of the scientific explosion. I shall show also,
from the statistical distribution, that it is reasonable on the
oretical grounds to suppose that the doubling time of one
measure might be a multiple of the period for some other index.
This treatment, however, requires a closer look at what is ac
tually being measured and must be deferred until further re
sults have been achieved from the study of the crude shape of
exponential growth.
Moreover, the “normal” law of growth that we have been con
sidering thus far describes, in fact, a most abnormal state of
events. In the real world things do not grow and grow until they
reach infinity. Rather, exponential growth eventually reaches
some limit, at which the process must slacken and stop before
reaching absurdity. This more realistic function is also well
From Derek J. de Sofia Price, Science Since Babylon (New Haven, Yale
known as the logistic curve, and it exists in several slightly University Press, 1961).
different mathematical forms. Again, at this stage of ignorance
of science in analysis, we are not particularly concerned with find asymmetrical logistic curves that use up one more param
the detailed mathematics or precise formulation of measure eter to describe them. Nature appears to be parsimonious with
ments. For the first approximation (or, more accurately, the her parameters of growth.
zeroth-order approximation) let it suflBce to consider the gen Because of the symmetry so often found in the logistic curves
eral trend of the growth. that describe the growth of organisms, natural and manmade,
The logistic curve is limited by a floor—that is, by the base measuring science or measuring the number of fruit flies in a
value of the index of growth, usually zero—and by a ceiling, bottle, the width of the curve can be simply defined. Mathe
which is the ultimate value of the growth beyond which it can matically, of course, the curve extends to infinity in both direc
22 A SC IE N C E OF SC IE N C E A SC IE N C E OF SC IE N C E 23
tions along the time axis. For convenience we measure the four days after the onset of the process, so that there is but one
width of the midregion cut off by the tangent at the point of day of relatively free growth, and final length is attained after
inflection, a quantity corresponding to the distance between the seven days. Note that the analysis involved no knowledge about
quartiles on a standard curve of error or its integral. This mid the height of the curve from floor to ceiling. True, we made a
region may be shown necessarily to extend on either side of statement about the date of the midpoint—it occurred after
the center for a distance equal to about three of the doubling four days of growth—but we could equally well have noted
periods of the exponential growth. that the exponential growth, short-lived in this case, extends
Thus, for example, if we have a beanstalk that doubles in only for the first day, and from this it would follow that three
height every day, there will exist a midperiod of about six days more doublings must bring it to the midpoint, and a further
during which the beanstalk will leave its juvenile exponential three to senescence.
growth and settle down to an adult life of stability in length Now, with no stronger assumption than has been made about
(Fig. 6). The only question is one of how much free and ex the previously regular exponential growth with a doubling
ponential growth is allowed before the decelerative period sets period of 10 to 15 years, we may deduce, as we have, that the
in. For the beanstalk, the midpoint of growth occurs only about existence of a ceiling is plausible since we should otherwise
reach absurd conditions at the end of another century. Given
the existence of such a limit, we must conclude that our ex
ponential growth is merely the beginning of a logistic curve
in other guise. Moreover, it is seen that as soon as one enters
the midregion near the inflection—that period of secession from
accustomed conditions of exponential growth—^then another 30
to 45 years will elapse before the exact midpoint between floor
and ceiling is reached. An equal period thereafter, the curve
will effectively have reached its limit. Thus, without reference
to the present state of affairs or any estimate of just when and
where the ceiling is to be imposed, it is apparent that over a
period of one human generation science will suffer a loss of its
traditional exponential growth and approach the critical point
marking its senile limit.
Fig. 6. GROWTH IN LENGTH OF A BEANSTALK AS A However, growths that have long been exponential seem not
FUNCTION OF AGE to relish the idea of being flattened. Before they reach a mid
point they begin to twist and turn, and, like impish spirits,
Adapted from D’Arcy W. Thompson, Growth and Form (Cambridge,
England, Cambridge University Press, 1948), p. 116, Fig. 20. change their shapes and definitions so as not to be exterminated
24 A SC IE N C E OF SC IE N C E A SC IE N C E OF SC IE N C E 25
against that terrible ceiling (Fig. 7). Or, in less anthropomor One therefore finds two variants of the traditional logistic
phic terms, the cybernetic phenomenon of hunting sets in and curve that are more frequent than the plain S-shaped ogive.
the curve begins to oscillate wildly. The newly felt constriction In both cases the variant sets in some time during the inflection,
produces restorative reaction, but the restored growth first presumably at a time when the privations of the loss of exponen
wildly overshoots the mark and then plunges to greater depths tial growth become unbearable. If a slight change of definition
than before. If the reaction is successful, its value usually seems of the thing that is being measured can be so allowed as to count
to lie in so transforming what is being measured that it takes a new phenomenon on equal terms with the old, the new logis
a new lease on Hfe and rises with a new vigor until, at last, it tic curve rises phoenixlike on the ashes of the old, a phenome
must meet its doom. non first adequately recognized by Holton and felicitously
called by him “escalation.” Alternatively, if the changed
conditions do not admit a new exponential growth, there will be
violent fluctuations persisting until the statistic becomes so ill-
defined as to be uncountable, or in some cases the fluctuations
decline logarithmically to a stable maximum. At times death
may even follow this attainment of maturity, so that instead of
a stable maximum there is a slow decline back to zero, or a
sudden change of definition making it impossible to measure
the index and terminating the curve abruptly in midair.
Logistic curves such as these have become well known in
numerous analyses of historical time-series, especially those
concerning the growth of science and technology. The plain
curve is well illustrated in the birth and death of railroad track
mileage; in this case the maximum is followed by an eventual
decline as tracks are tom up and lines closed down. The curve
lo ^seiCCaUoK CcKverffgfd osciCltUiofo followed by hunting fluctuations appears in the flgures for the
Fig. 7. WAYS IN WHICH LOGISTIC GROWTH MAY production of such technological raw materials as coal and
REACT TO CEILING CONDITIONS metals (Fig. 8).® The escalated curves are probably the most
In escalation, new logistics are bom as the old ones die, in loss of common and can be seen in the number of universities founded;
definition it becomes impossible to continue to measure the variable the separate steps here beautifully reflect the different tradi-
in the same way or in the same units, and in oscillation ( convergent and
divergent) cybernetic forces attempt to restore free growth. • S. G. Lasky, “Mineral industry futures can be predicted,” Engineering
and Mining Journal, 152 (August, 1951), 60; 156 (September, 1955), 94.
A SC IE N C E OF SC IE N C E 27
tions of the medieval universities and the Renaissance founda
tions (Fig. 9).
They can be seen again in the now-famihar graph, first pre
sented humorously by Fermi,^ showing the power of acceler-
Numbcr of
From the foundation at Cairo in 950 up to ca. 1460 there is pure ex
ponential growth, doubling in about 100 years. Thereafter saturation
sets in, so that the midregion of the sigmoid extends from 1300 to
ca. 1610. Between 1460 and 1610 is a period of transition to the new
form of universities, a growth that also proceeds exponentially as if it
had started from unity ca. 1450 and doubling every 66 years. There is
probably an ever-greater transition to yet faster growth starting at the
end of the Industrial Revolution. From Derek J. de Solla Price, Science
Since Babylon (New Haven, Yale University Press, 1961).
’ The exponential growth of accelerators was first noted by John P.
Blewett in an Internal Report of the Cosmotron Department of Brook-
haven National Laboratory, written on June 9, 1950. The first public
Fig. 8. LCXJISTIC GROWTH O F RAW M ATERIAL PRODUCTION, presentation of this material was made by Fermi in his address as retir
SHOWING OSCILLATION ON ATTAINING CEILING CONDITIONS ing President, at the American Physical Society meeting in January, 1954.
Figure 10 is a later version from M. S. Livingston and J. P. Blewett,
Adapted from S. G. Lasky, “Mineral industry futures can be predicted,* Particle Accelerators (New York, McGraw-Hill Book Company, Inc.,
Engineering and Mining Journal, 156 (September, 1955). 1962), p. 6.
)0-
0 ------------------
/ i
® ; A It /
/ i
0 ::----------------- >
3:
4 c _ .
—
/ /
/
/ '*
/
;
A
-A
/< /'
— 1
'A : /
/
r
/ I
V ji
■ 1.11.1 - L .J l,.l . - 1 . 1 .1 .J _ i _L_1. I 1 . 1 1 11 1- » ■ 1
Fig. 11. NUM BER OF CHEM ICAL ELEM EN TS KNOWN
AS A FUNCTION OF DATE
Fig. 10. THE RATE OF INCREASE OF OPERATING ENERGY
After the work of Davy there is a clear logistic decline followed by a
IN PARTICLE ACCELERATORS
set of escalations corresponding to the discovery of elements by tech
niques that are predominantly physical. Around 1950 is the latest es
From M. S. Livingston and J. P. Blewett, Particle Accelerators (New
calation produced by the manufacture of trans-uranic elements.
York, McGraw-Hill Book Company, Inc., 1962), p. 6, Fig. 1.1, used by
permission.
30 A SC IE N C E OF SC IE N C E A SC IE N C E OF SC IE N C E 31
ators (Fig. 10). It becomes less and less humorous as it goes on nature of this change, and any interpretation of it, must depend
faithfully predicting when yet another major advance in method on what we are measuring and on how such an index is com
is needed to produce another step in the escalation. Yet, again, piled.
escalations can be seen in the curve showing the number of Even without such definition and analysis one can immedi
chemical elements known as a function of date (Fig. 11). ately deduce various characteristics of such a period. Clearly
Omitting the jfirst ten, which were known to prehistoric man, there will be rapidly increasing concern over those problems of
we have a steady exponential growth, doubling a shade more manpower, literature, and expenditure, that demand solution
rapidly than every 20 years, followed by a midpoint in about by reorganization. Further, such changes as are successful will
1807 when Sir Humphrey Davy had his heyday, then a period lead to a fresh escalation of rapid adaptation and growth.
of decline when the first 60 elements had been found. By the Changes not efiBcient or radical enough to cause such an off
end of the nineteenth century, when new methods, physical shoot will lead to a hunting, producing violent fluctuations that
rather than chemical, led to new classes of elements, there ap will perhaps smooth out at last.
peared a new bunch of ogives, then a halt until the big ma Such an analysis seems to imply that the state called Big Sci
chines enabled man to create the last batch of highly unstable ence actually marks the onset of those new conditions that will
and short-lived trans-uranic elements. break the tradition of centuries and give rise to new escalations,
From this we are led to suggest a second basic law of the violent huntings, redefinitions of our basic terms, and all the
analysis of science; all the apparently exponential laws of other phenomena associated with the upper limit. I will suggest
growth must ultimately be logistic, and this implies a period of that at some time, undetermined as yet but probably during the
crisis extending on either side of the date of midpoint for about 1940s or 1950s, we passed through the midperiod in general
a generation. The outcome of the battle at the point of no return logistic growth of science’s body politic.
is complete reorganization or violent fluctuation or death of Thus, although we recognize from our discussion so far that
the variable. saturation is ultimately inevitable, it is far too approximate to
Now that we know something about the pathological after indicate when and in what circumstances saturation will begin.
life of a logistic curve, and that such things occur in practice in We now maintain that it may already have arrived. It may seem
several special branches of science and technology, let us re odd to suggest this when we have used only a few percent of
open the question of the growth curve of science as a whole. We the manpower and money of the country, but in the next chap
have seen that it has had an extraordinarily long life of purely ter it will appear that this few percent actually represents an
exponential growth and that at some time this must begin to approach to saturation and an exhausting of our resources that
break down and be followed by a generation-long interval of in nearly (within a factor of two) scrapes the bottom of the barrel.
creasing restraint which may tauten its sinews for a jump either At all events, the appearance of new phenomena in the in
toward escalation or toward violent fluctuation. The detailed volvement of science with society seems to indicate something
32 A SC IE N C E OF SC IE N C E
radically different from the steady growth characteristic of the
entire historic past. The new era shows all the familiar syn
dromes of saturation. This, I must add, is a counsel of hope
rather than despair. Saturation seldom implies death, but rather
(2)
that we have the beginning of new and exciting tactics for sci
ence, operating with quite new ground rules. GALTON
It is, however, a grave business, for Big Sci^ c e inteqireted- REV ISI TED
tlius becomes an uncomfortably brief interlude between the
traditional centuries of Little Science and the impending period
following transition. If we expect to discourse in scientific style
about science, and to plan accordingly, we shall have to call this
approaching period New Science, or Stable Saturation; if we FRANCIS GALTON (1822-1911), grandson of Erasmus Darwin,
have no such hopes, we must call it senility. was one of the most versatile and curious minds of the nine
teenth century. He brought fingerprinting to Scotland Yard,
founded the Eugenic Society which advocated breeding of the
human race on rational principles, and, above all, gave a flying
start to the science of mathematical genetics. His passion was
to count everything and reduce it to statistics. Those who see
the social sciences rising on a solid foundation of quantified
measurements and mathematical theory might wefl take him as
a patron saint rather than Sir William Petty, who is usually seen
as the first to bring numbers into the study of people by ana
lyzing the bills of mortality in the seventeenth century.
Galton’s passion shows itself best, I feel, in two essays that
may seem more frivolous to us than they did to him. In the first,
he computed the additional years of life enjoyed by the Royal
Family and the clergy because of the prayers offered up for
them by the greater part of the population; the result was a neg
ative number. In the second, to relieve the tedium of sitting for
a portrait painter, on two different occasions he computed the
number of brush strokes and found about 20,000 to the portrait;
34 GALTON R E V ISIT E D GALTON R E V ISIT E D 35
just the same number, he calculated, as the hand movements telling us how many men or scientific papers or pieces of re
that went into the knitting of a pair of socks.^ search there are at each of several levels of quality, is necessary
Let it not be thought that Galton was some sort of crank. His if we are to understand the natmre of scientific quality, and this
serious work was of the highest standards of scholarship and knowledge is a prerequisite to the interpretation of the several
importance, but he is now increasingly neglected because, al different index measures previously mentioned in connection
though his researches were founded on the exciting and valid with the basic laws governing the exponential and logistic rates
basis of Darwin s theory of evolution, Galton had missed the of growth of science. The second will help us formulate ground
true mechanism of genetic action, discovered by his exact con rules for what to expect of scientists when the change of condi
temporary, Mendel. Mendel published his findings just five tions produced by Big Science or Saturation Science alters their
years before Galton's book on hereditary genius,^ but was not circumstances from those they had known in past ages.
discovered by the outside world until Galton was nearly 80. Galton began by estimating how rare in the England of his
We shall examine his book Hereditary Genius, and, with par day were various types of men who were engaged in human
ticular attention, his special study, English Men of Science affairs generally and in science particularly and who were of
(London, 1874). In these works Galton is primarily concerned sundry degrees of eminence. Using the criterion that a man was
with his thesis that great men, including creative scientists, tend eminent if his name appeared in a short biographical compila
to be related and that therefore a series of elite families con tion of 2500 Men of the Time that had just been published, or
tributed perhaps the majority of distinguished statesmen, sci in the select columns of obituary notices in The Times, he found
entists, poets, judges, and military commanders, of his day and that such noteworthiness had an incidence of about one person
of the past. His main work is full of pitfalls, and currently we for every 2000 adult males or one person in 20,000 of the general
are not concerned so much with the Galtonian approach to population—a mere handful alive at any time in the country.
genetics as we are with several of his interesting side investiga For eminent scientists he set a standard which demanded
tions. These are his pioneer studies of the distribution of quality that they should be not merely Fellows of the Royal Society—a
among distinguished scientists, and a set of summaries that we meaningful honor since the reforms of election under Mr. Jus
should nowadays call sociological and psychological, telling us tice Groves some 30 years before—^but that they must be further
something about the characteristics of these exceptional men. distinguished by a university chair; by a medal presented by a
We intend to review these two main lines in the light of the learned society, or an office held in such a body; or by member
twentieth century and its extensions of Gabon’s work. The first, ship in some elite scientific club of academic worth. His count
of people from whom he could obtain the full biographical in
* Karl Pearson, The Life, Letters and Labours of Francis Galton ( New
York, Cambridge University Press, 1914-30), see especially Vol. Ilia, formation desired was 180, and he estimated that in the entire
p. 125. country there might be at the most 300 such people.
* Francis Galton, Hereditary Genius (London, 1869; reprinted by
Meridian Books, 1962), Reckoning that half of them were between the ages of 50
36 G ALTO N R E V ISIT E D GALTON R E V ISIT E D 37
and 65, he calculated that the chance of rising to such stature names to a million population of the United States, in the vol
was about 1 in 10,000 adult males of this age group, a figure ume for 1938 there are about 12.4 to a million, and that both
roughly corresponding to 1 in 100,000 of the general popula figures are of the same order of magnitude as that found by
tion. However, since the general biographical lists show that Galton. Certainly there appears to have been no vast change in
only about 1 in 10 eminent men was engaged in science or medi the number of “eminent” men of science to a million population,
cine, then by his previous standards there should only have either on moving our scene of inquiry from Britain to the
been about one eminent scientist for every 200,000 of the gen United States or on following it through nearly a doubling of
eral population. The fact that Galton supposes there to have the United States population. One may argue that Galtons
been twice as many, means either that he was erring on the side standard of distinction is not the same as Cattell’s. One may
of generosity in estimating the numbers of good scientists who maintain with even greater reason that Cattell’s arbitrary allo
should have been on his list and were not, or that the tendency cation of a set quota of 1000 stars originally, with a fixed incre
is to cast a broader net when looking for great scientists than ment thereafter, was perhaps out of all proportion to a constant
when looking for great men in general. standard of eminence. In spite of this, we can find no rapid
The utility of this investigation is that it provides an estimate changes in this estimated incidence of scientific eminence.
of the number of scientific persons whom Galton considered If in studying American Men of Science we look not at the
important enough to be well worth discussion, but without starred names alone but at all of them, we observe a most
limiting the scope to include such a small group that it would strildng change in order of magnitude with the passing of time
leave the investigator generalizing about a mere handful of ( Table 1). Just to run one’s eye along the set of 10 editions on
geniuses. Thus, between 5 and 10 persons in a million fall within a shelf is to feel an immediate respect for the power of exponen
this category. How does this compare with the state of affairs
tial growth.
since Galton s time? It is apparent that within the past 50 years there has been a
Fortunately there is an admirable biographical compilation, sixteenfold increase in the number of men, an exponential
American Men of Science, that has run through 10 editions growth with a doubling period of about 1234 years, a figure
between 1903 and 1960. The editor, J. McKeen Cattell (himself already suggested as typifying the growth of science. Even in
a prominent psychologist), rendered signal service by starring relation to the size of the general population it can be seen that
the most noteworthy names, beginning with an original 1000 the same half-century has multiplied the density of scientists by
and adding to this number as each new edition appeared.^ It a factor of eight, a doubling in about 17 years. Another four
so happens that in the first edition there are about 11 starred such half-centuries of regular growth would give us more than
* Here and later we have made considerable use of the extensive analy two million American men of science per million population,
sis in S. S. Visher, Scientists Starred 1903-1943 in “American Men of if it were not that exponential growths inevitably become logis
Science," (Baltimore, The Johns Hopkins Press, 1947).
tic and die.
3 8 G A LT O N R E V IS IT E D GALTON R E V ISIT E D 39
We have already shown that because of this logistic ma ard extent. The same problem is encountered in most recent
chinery the prospect for the immediate future is more interest evaluations of the high-talent population on the basis of intelli
ing than that of a slow death from suffocation in a .d . 2160. Our gence tests. For example, one may say that on an AGCT (Army
crisis seems to be but a few decades ahead, and far more in General Classification Test) type of test only one man in 10,000
volved with the nature of the growth than with the final ex of his age group might score more than 170, one in 100,000 more
haustion of the population. It is therefore a matter of some than 180, one in 1,000,000 more than 190, and so on, with an or
interest to seek the reason why, in spite of this general rapid der of magnitude for each 10 (more accurately, 11) points that
raise the stakes. But one cannot usefully say that eminence be
TABLE 1 gins at a score of 172 and not below. Even if genius were merely
NUMBER OF MEN CITED IN EDITIONS OF a matter of the talents being measured by the test in hand,
AMERICAN MEN OF SCIENCE there would be no clear cutoff, only a gradual falling off of the
population as the standards are raised. The fault lies not so
Year of publication Number of men population much in the definition of what constitutes scientific ability as
1903 4,000 50 in the false premise that distinction or genius can be decided
1910 5,500 60
1921 on a yes-or-no-basis.
9,500 90
1928 13,500 Results more accurate, although not much more, can be
110
1933 22,000 175 achieved by taking a reasonably small group of tabulated men,
1938 28,000 220 discoveries, or even scientific institutions, journals, and coun
1944 34,000 240
1948 tries, and carefully marking in some special way those that were
50,000 340
1955 74,000 440 distinguished. For eminent men, for example, one might use as
1960 (omitting social 96,000 480 criteria selection to give invited papers, and to receive medals
sciences) and other awards, such as Nobel Prizes. This gives the usual
sort of exponential growth, but with a doubling time consider
exponential growth of scientific manpower—and, incidentally, ably longer than 10 years. For example, in a select, apparently
of its publications and budgets—the number of truly great men superior group of modern scientists in any large field, drawn
does not seem to change with the same quick exuberance. from standard biographical handbooks or other sources that
The root of the trouble, as Galton well perceived, lies in the select only a small elite, the doubling time is about 20 years.
establishment of any objective standard of eminence not de One obtains about the same figure for any list of selected great
pendent upon time. Conceivably, all that we have said is that scientific discoveries.
when men are chosen by degrees of selectivity that run to To improve the strength and significance of this result, it is
orders of 10 in a million they become remarkable to this stand clearly necessary to make some statement about degrees of
40 GALTON R E V ISIT E D G A LTO N R E V ISIT E D 41
eminence that would give not a dichotomy of distinguished and honored. Conversely, the low-scoring end of the list contains
undistinguished but rather a sliding scale, a sort of velocity dis fewer such names in terms of absolute numbers, and much
tribution. One such scale—the traditional one used by deans fewer in proportion.
and other employers as a measure of scientific success—^is the Exactly such a study was made by Wayne Dennis. Using as
number of publications produced by each man in accepted his source the National Academy of Sciences Biographical
scientific journals. Let it be freely admitted at the outset that Memoirs for 1943-52, he showed that of the 41 men who died
this is a bad scale. Who dares to balance one paper of Einstein after a full life, having reached the age of 70, the top man had
on relativity against even a hundred papers by John Doe, 768 publications, the bottom 27. The average number of publi
Ph.D., on the elastic constant of the various timbers (one to a cations was more than 200, and only 15 persons had fewer than
paper) of the forests of Lower Basutoland? 100 in their bibliographies. Similarly, a list of 25 eminent
The scale is bad if for no other reason than that its existence nineteenth-century scientists showed that all but one were in
has moved people to publish merely because this is how they the range of 61 to 307 items.^ Further, taking a sample from the
may be judged. Nevertheless, it makes a starting point, and Royal Society Bibliography of Scientific Literature 1800-1900,
later on it may be refined to meet objections. We shall show, for he showed that the most productive 10 percent of all authors,
example, that all such distributions are of the same type and, having each more than 50 publications, were of such caliber
thus, though one cannot directly measure “scientific ability,” that 50 percent of them gained the distinction of mention in the
one may reasonably deduce properties of its presumed distribu Encyclopaedia Britannica; of the top 5 percent, each of whom
tion. We shall also have to enter the caveat that the scale may had more than 140 items to his bibliography, some 70 percent
not be directly applicable to the era of Big Science, which has received such mention. None of those mentioned in the Ency-
involved so much collaborative work that one cannot easily de clopaedia by virtue of their scientific work had fewer than seven
termine a man’s score. This is another point to be reserved for publications.
later elaboration. Thus, although there is no guarantee that the small producer
Let us not begin with too pessimistic an outlook on the worth is a nonentity and the big producer a distinguished scientist, or
of this investigation. Flagrant violations there may be, but on even that the order of merit follows the order of productivity,
the whole there is, whether we like it or not, a reasonably good there is a strong correlation,® and we are interested in looking
correlation between the eminence of a scientist and his pro deeper into the relative distribution of big- and small-output
ductivity of papers. It takes persistence and perseverance to be writers of scientific literature. Such studies are easy to make by
a good scientist, and these are frequently reflected in a sus counting the number of items under each author’s name in
tained production of scholarly writing. Then, again, it may be ‘ The exception being Riemann, who published only 19 papers but
well demonstrated that the list of high scorers contains a large died at the age of 40.
“ Productivity is therefore one of many factors.
proportion of names that are not only well known but even
GALTON R E V ISIT E D 43
the cumulative index of a journal. A pioneer investigation of
this sort was made by Lotka,® and several others have since re
peated such head counts. They all confirm a simple, basic result
that does not seem to depend upon the type of science or the
date of the index volume; the only requirement is that the index
extend over a number of years sufficient to enable those who
can produce more than a couple of papers to do so.
The result of this investigation is an inverse-square law of
productivity (Fig. 12). The number of people producing n
papers is proportional to 1/n^. For every 100 authors who pro
duce but a single paper in a certain period, there are 25 with
two, 11 with three, and so on. Putting it a httle diflFerently by
permitting the results to cumulate, one achieves an integration
that gives approximately an inverse first-power law for the num
ber of people who produce more than n papers; thus, about one
in five authors produces five papers or more, and one in ten
produces at least ten papers (Fig. 13).
It is surprising that such a simple law should be followed so
accurately and that one should find the same distribution of
scientific productivity in the early volumes of the Royal Society
as in data from the twentieth-century Chemical Abstracts. The
regularity, I suggest, tells us something about the nature of the
scores we are keeping. An inverse-square law probability dis
tribution, or an inverse first power for the cumulative prob
ability, is nothing like either the normal Gaussian or Poisson
distributions, or any of the other such curves given by normal
hnear measure of events that go by chance. If the number of
Fig. 12. LOTKA S LAW scientific papers were distributed in a manner similar to that of
* Alfred J. Lotka, “The frequency distribution of scientific productivity,”
The number of authors publishing exactly n papers, as a function of n.
Journal of the Washington Academy of Sciences, 16 (1926), 317. For
The open circles represent data taken from the first index volume of the
a fuller analysis and justification see Herbert A. Simon, Models of Man,
abridged Philosophical Transactions of the Royal Society of London
Social and National (New York, John Wiley and Sons, Inc., 1957), p. 160.
(17th and early 18th centuries), the filled circles those from the 1907-16
decennial index of Chemical Abstracts. The straight line shows the exact
inverse-square law of Lotka. All data are reduced to a basis of exactly
100 authors publishing but a single paper.
G A L T O N R E V IS IT E D 4 5
TABLE 2
SCHEMATIC TABLE SHOWING NUMBERS OF AUTHORS
OF VARIOUS DEGREES OF PRODUCTIVITY (IN
PAPERS PER LIFETIME) AND NUMBERS
OF PAPERS SO PRODUCED‘‘
Papers/man Men Papers
1 100 100 (The 75 percent of men who
2 25 50 are low scorers produce one-
3 11.1 33.3 quarter of all papers.)
4 6.2 25
5 4 20
6 2.8 16.7
7 2 14.2
8 1.5 12.5
9 1.2 11.1
10 1 10
10-11.1 1 10-f
11.1-12.5 1 11.1-1-
12.5-14.2 1 12.5+
14.2-16.7 1 14.2+ (Subtotal: 10 men produce
16.7-20 1 16.7+ more than 50 percent of all
20-25 1 20+ papers.)
25-33.3 1 25+
33.3-50 1 33.3+
50-100 1 50+ (The top two men produce
Over 100 1 100+ one-quarter of all papers.)
Total 165 586+
Average papers/man = 586/165 = 3.54
• Table constructed on basis of exactly 100 men with a single pub
Fig. 13. NUM BER OF AUTHORS PUBLISHING AT LEA ST fl PAPERS lished paper. Other entries computed from Lotka’s law.
AS A FUNCTION OF fl
Same data, and same reduction as for Fig. 12, but full curve here is
modified to a form that takes account of Lotka’s overestimation of the
number of highly prolific authors (see footnote 8, Chapter 2 ).
46 GALTON R E V ISIT E D
have the easier it seems to be to get the next, a principle to
Nvhich we shall return later.
Let us first examine the nature of the crude inverse-square
law of productivity (Table 3). If one computes the total pro
duction of those who write n papers, it emerges that the large
number of low producers account for about as much of the
total as the small number of large producers; in a simple sche
matic case, symmetry may be shown to a point corresponding
to the square root of the total number of men, or the score of
the highest producer. If there are 100 authors, and the most
prolific has a score of 100 papers, half of all the papers will have
been written by the 10 highest scorers, and the other half by
those with fewer than 10 papers each. In fact, in this ideal case,
a full quarter of the papers have been written by the top two
men, and another quarter by those who publish only one or
two items.
This immediately gives an objective method for separating
the major from the minor contributors. One may set a limit and
say that half the work is done by those with more than 10 papers
to their credit, or that the number of high producers seems to
be the same order of magnitude as the square root of the total
number of authors. The first way, setting some quota of 10 or
✓ <
so papers, which may be termed “Deans’ method,” is familiar
Fig. 14. NUM BERS OF PUBLICATIONS OF FOUR SERIES OF HIGHLY
enough; the second way, suggesting that the number of men
DISTINGUISHED AND ( IN Cm EN TA LLY) HIGHLY PROLIFIC
goes up as the square of the number of good ones, seems con
AUTHORS, EACH RANKED W ITHIN TH E SERIES
sistent with the previous findings that the number of scientists
doubles every 10 years, but the number of noteworthy scientists The series are (1 ) members of the National Academy of Sciences, drawn
from obituary bibliographies, (2 ) nineteen eminent scientists of the 19th
only every 20 years. century, (3 ) most prolific authors in decennial index of Chemical Ab~
Unfortunately, Lotka’s simple inverse-square law needs stracts, (4 ) index to Vols. 1-70, Philosophical Transactions of the Royal
modification in the case of high scorers ( Fig. 14). Beyond the Society.
of the United States crop of Ph.D.’s for 1958 enable us to say telligence recorded, AGCT 170+, about one person in five re
something of the incidence and of the intelligence-test char ceived a Ph.D., although the general incidence of doctorates in
acteristics of this group. Now, the Ph.D. and the editorial stand the age group was only 1 in 3000. Thus, intelhgence has a lot
ards of learned periodical publications are things that we have to do with the gaining of Ph.D.’s. If we now consider it plau
done our best to keep constant. It is therefore reasonable to sible that this current figure of one in five refers to those supe
identify the minimum effort of writing a single scientific paper rior beings who become highly productive scientists, one could
with that demanded by the “sheepskin gateway” to the road of contemplate using all means, fair and foul, to close the gap so
research. Although it is agreed that these things do not coincide, that they would all earn Ph.D.’s or even scientific Ph.D.’s.
since some Ph.D.’s never publish even their theses, whereas We know now that the total number of scientists goes up as
many authors are not doctors, yet at worst they should differ by the square, more or less, of the number of good ones. There
some reasonably constant ratio not too far from unity. fore, if we want to multiply the good scientists by five, we must
Harmon found that in an age group of the population num multiply the whole group by 25. Instead of an age group of
bering about 2,400,000 there arises an annual crop of about about 8000 Ph.D.’s in mixed subjects, we should then have
8000 Ph.D.’s in all fields, the physical and biological sciences about 200,000, all in science. As it happens, the intelligence dis
together comprising about half the total. As one might expect, tribution shows that in an age group of 2,400,000, a few more
than 160,000 achieve AGCT 130, and so we have a minimal
the intelligence-test scores for this group were considerably
higher than the general level, the average being AGCT 130*8 cutoff for possible scientists that is only slightly less than the
for the mode of the distribution. Taken by fields, there was a present mode found for Ph.D.’s, both scientific and otherwise.
variation from 140*3 for physics to 123*3 for Ph.D.’s in educa The two methods thus coincide to indicate that about 6 to 8
tion: percent of the population at most could be minimal scien
tists.
Physics 140*3
Mathematics 138*2 Apparently, then, the scale of solidness in scientific publica
Engineering 134*8 tion should have its zero placed at an AGCT level of about 130,
Geology 133*3 corresponding to about one person in every 15 in an age group.
Arts and humanities 132*1 Attractive though it may be to perceive such a cutoff point,
Social sciences 132*0
Natural sciences 131*7 agreeing as it does so well with the present norm for Ph.D.’s,
Chemistry 131*5 the impHcations are grave. At first sight it appears that at pres
Biology 126*1 ent we are tapping only about one in 25 of those who could be
Education 123*3
come scientists at all, and a fifth of those who would be out
When these data were applied to the general population in standing scientists. If we took all the talent of the population
the same age group, it appeared that at the highest level of in with no loss or wastage, we should then have 8,000,000 scien-
5 4 G A L T O N R E V IS IT E D GALTON R E V ISIT E D 55
tists writing papers in the United S t a t e s , ^nd, of these, 80,000 I think we have now laid the theoretical basis for this study
would be highly productive, with more than 10 papers each. of science. It is remarkably similar to the study of econometrics.
Thus, we should have a roll of 40,000 scientists to a million On the one hand we have the dynamic treatment that gives us
population, and, of these, 400 in a million would be men of note. time series, first of exponential growth, then of the saturated
Galton, you remember, found about 5 to 10 eminent scientists growth resulting in standard logistic curves. On the other hand,
in a million population, and the early volumes of American Men we have the statics of a distribution law similar to that of
of Science showed 50 in a million. Thus, in the density of good Pareto. The extent of the difference between analyzing science
scientists we have left one more order of magnitude at the most and analyzing business hes in the parameters. The main ex
and, even at the expense of all other high-talent occupations, ponential part of the growth of science doubles in 10 years only,
science is not likely to engross more than 8 percent of the popu which is much more rapid than all else; the characteristic index
lation. Even so, it looks as if the decreasing return of good sci of the distribution law is one at the low end and two at the
entists to every 100 Ph.D.’s will make it more and more difficult high, instead of a uniform 1*5.
to reach a level of this magnitude. Just how strong is this hmi- The additional contributions that we have made lie in pro
tation? Is it possible that the level of good scientists cannot rise viding a reasonable theoretical basis for our Pareto law and in
by the factor of five that we have presumed? showing that, although the average number of papers per
Almost half of the factor is accounted for by the wastage of author remains sensibly constant, one may make a split between
scientific womanpower, a wastage that the U.S.S.R. has par those whose productivity is high and that much larger mass of
tially checked but that we seem unable to avoid. Another factor authors whose productivity is low. This mass is seen to grow as
of two might be attributed to the lack of opportunity and in the square of the number of high scorers, and therefore the
centive in regions outside the big cities where schools are good number of high scorers will appear to double only every 20
and competition and inspiration keen. Indeed, all things con years.
sidered, the high proportion of talented manpower successfully The Fechner law principle which we invoked to reduce the
diverted into science at present is surely to our credit. But if Pareto-hke distribution to the sort of linear and additive meas
the level cannot indeed rise, then we are, as we have already ure that is necessary for a standard probability curve is much
conjectured, about halfway toward saturation at the top end of more powerful than we have yet assumed. If we may take in
the scale, and any increase in numbers of scientists must pro general the solidness of a body of pubhcations as measured by
duce an even greater preponderance of manpower able to write the logarithm of the number of papers, it has further interesting
scientific papers, but not able to write distinguished ones. It consequences. Consider the law of exponential growth previ
gives serious pause to thoughts about the future of scientific ously mentioned as a universal condition of freely expanding
education. Is it worth much sacrifice? science. Obviously, the solidness of the field, the logarithm of
“ This is more than twice the present world population of scientists.
the number of papers, grows linearly with time. Thus, since it
56 G A LTO N R E V ISIT E D G A LTO N R E V ISIT E D 57
takes about 50 years for the number of men or number of papers ural growth enabling them in general to maintain their lead. It
in a field to multiply by 10, there is a unit increase of soHdness is the exception, rather than the rule, for one of the big blocks
every half-century.^® to slacken its growth—presumably through the existence of
I cannot quite see why it is so, or how one might judge it other some sort of logistic ceiling that causes it to stagnate—and be
than by pure intuition, but the two units of solidness separating overtaken so that it falls in rank.
the man who can publish no more than one paper in a lifetime The fact that the general growth of science increases equally
from the one who can write a hundred such papers are essen the sizes of the large blocks and the numbers of the small blocks,
tially the same as those that separate the two states of a subject while presenting an appearance of crystallization, is really not
at dates a century apart. In rough, and misleading, terms one so peculiar. Precisely the same thing happens when the popu
might say that the eminent scientist is a century ahead of the lation of a country grows. Instead of being uniformly distrib
minimal one. uted over the country, it is crystallized out into variously sized
What further implications are there of the assumption that blocks called cities. The growth of cities in a country provides a
one can measure the progress of a field by the linear march of useful model for the growths of scientific blocks within science.
its solidness? Are such degrees of solidness truly additive? Must As it happens, the hierarchical order of cities or other blocks,
we judge one field of a hundred workers adding two units of ranked by decreasing size, offers yet another example of the
solidness within a certain time as inferior to 10 separate fields same Pareto-like distribution we have already found for the
of 10 workers, each of whom will add one unit to each field, productivity of scientific authors.
making a total of 10 units within the same time? In the case of cities, the historical statistics provide a good
If such an indication be true, then it seems that science has example of such a distribution on the move, with everything
a strong desire to minimize its solidness rather than make it as increasing exponentially while maintaining the normal distribu
large as possible. Beyond the phenomenon of exponential tion (Fig. 15).^^ Using a plot showing the distribution at each
growth, science displays in several ways a tendency to crystal decade, one may see the constant slope of the distribution on a
lize out, in the sense that big things grow at the expense of the log-log scale and the inexorable march of the intercepts that
small ones that constitute a sort of mother liquor. Large fields tell us the magnitude of the biggest city on the one scale and
seem to absorb the manpower and subject matter of small ones. the number of minimal cities (here taken as population 2500)
Even though new fields, new departments, new institutions, on the other scale. Both increase regularly each decade, taking
and even new countries arrive on the scientific scene in increas about 60 years each to go through a power of 10 or, as we have
ing number, the few previously existing large ones have a nat- called it before, one unit of solidness. If one looked in detail at
“ This, then, provides a measure that is linear, not exponential. It is
Figure 15 and the following data are from G, K. Zipf, Human Be
the sort of index which might correspond with Nobel Prizes (which
havior and the Principle of Least Effort (Cambridge, Mass., Addison-
come linearly with time because that is how they are organized); pos-
>iblv also with unexpected, crucial advances. Wesley Publishing Company, Inc., 1949), p. 420, Fig. 10-2.
GALTON R E V ISIT E D 59
well by such diverse hierarchical hsts as those giving the sizes
in faculties, or in Ph.D.’s per decade, of the college scientific
departments, in any field or in general, in the United States or
in the world. It is followed by ranked lists showing the scientific
contributions, in terms of papers, journals, or expenditures of
the nations of the world, ranging from the few big producers
on any scale relative or absolute to the minor production of the
large number of underdeveloped countries (Fig. 16).^®
About this process there is the same sort of essential, built-in
undemocracy that gives us a nation of cities rather than a coun
try steadily approximating a state of uniform population den
sity. Scientists tend to congregate in fields, in institutions, in
countries, and in the use of certain journals. They do not spread
out uniformly, however desirable that may or may not be. In
particular, the growth is such as to keep relatively constant the
balance between the few giants and the mass of pygmies. The
number of giants grows so much more slowly than the entire
population that there must be more and more pygmies per giant,
deploring their own lack of stature and wondering why it is that
neither man nor nature pushes us toward egalitarian uni
formity.
Value judgments aside, it seems clear that the existence of a
reasonable distribution that tells us how many men, papers,
countries, or journals there are in each rank of productivity,
utility, or whatever you will measure provides a powerful tool.
Communities of 2500 or more inhabitants ranked in the decreasing order Instead of attempting to get precision in defining which heads
of population size. It should be noted that the distribution at any given
date shows size decreasing uniformly with rank; as cities become more to count in exponential growth, one may instead take a crude
numerous and all of them increase in size, the distribution pattern is count and interpret it by means of such a distribution.
preserved, the curve moving parallel to itself at a constant rate. From Just as one cannot measure the individual velocities of aU
George K. Zipf, Human Behavior and the Principle of Least Effort { Cam
bridge, Mass,, Addison-Wesley Publishing Company, Inc., 1949), p. 420, ’®Data from a preliminary survey of scientific periodicals by the Li
Fig. 10-2.
brary of Congress.
G A LTO N R E V ISIT E D 61
molecules in a gas, one cannot actually measure the degrees of
eminence of all scientists. However, there are reasonable
grounds for saying that such measurements, if made, would
follow the standard distribution. In particular, we can take this
Pareto-like distribution as a hypothesis and see how the con
sequences agree with gross phenomena which we can measure.
We do, in fact, find a reassuring agreement.
Such, then, is the broad mathematical matrix of exponential
growth, logistic decay, and distribution functions. It provides
us now with a general description of the normal expansion of
science and its state at any time. Knowing now the regular
behavior, we have a powerful tool for investigating the signifi
cant irregularities injected into the system by the gross per
turbations of war and revolution, by the logistic birth and death
of measurable entities, by genius and crucial discovery, and,
in short, by all the organizational changes within the body
politic of science and in its relations with the state and society
in general.
IN V IS IB L E C O L L E G E S 63
because there were too many books. Here is a cry from the
heart of a scholar:
meet in the “Rochester Conference” for fundamental particles find a closed group, a small number of hundreds in member
studies, and in the similar number who congregate by invita ship strength, selected from a population of a large number
tion to discuss various aspects of solid state physics. of tens of thousands.
The organization is not perfect; a few of the best men may In addition to the mailing of preprints, ways and means
not attend, a few of those who do attend might not qualify are being found for physical juxtaposition of the members.
if we had perfect objective judgment. Conscientiously, one They seem to have mastered the art of attracting invitations
might try not to be too exclusive, not to bar the gentleman from centers where they can work along with several members
from Baffinland who would be a distinguished researcher on of the group for a short time. This done, they move on to the
fundamental particles if only he could. But there is a limit next center and other members. Then they return to home
to the useful size, and, if too many are invited, an unoflBcial base, but always their allegiance is to the group rather than
subgroup of really knowledgeable members will be forced into to the institution which supports them, unless it happens to
being. be a station on such a circuit. For each group there exists a
Such activity is by no means confined to the two groups sort of commuting circuit of institutions, research centers,
mentioned. Similar unoflBcial organizations exist in molecular and summer schools giving them an opportunity to meet piece
biology, in computer theory, in radio astronomy, and doubtless meal, so that over an interval of a few years everybody who is
in all sciences with tens of thousands of participants. By our anybody has worked with everybody else in the same category.
theory they are inevitable, and not just a product of the war Such groups constitute an invisible college, in the same
or the special character of each discipline. Conferences are sense as did those first unoflBcial pioneers who later banded
just one symptom; it becomes insuflBcient to meet as a body together to found the Royal Society in 1660. In exactly the
every year, and there is a need for a more continuous means same way, they give each man status in the form of approba
of close contact with the group of a hundred. tion from his peers, they confer prestige, and, above all, they
And so these groups devise mechanisms for day-to-day com eflFectively solve a communication crisis by reducing a large
munication. There is an elaborate apparatus for sending out group to a small select one of the maximum size that can be
not merely reprints of publications but preprints and pre handled by interpersonal relationships. Such groups are to be
preprints of work in progress and results about to be achieved.^* encouraged, for they give status pay-oflE without increasing
The existence of such a group might be diagnosed by check the papers that would otherwise be written to this end. I
ing the preprint list of one man and following this by a check think one must admit.that high-grade scientific commuting
of the list of each man mentioned. I think one would soon has become an important channel of communication, and that
we must ease its progress.
“ Like government contract research reports, these represent an ob
noxious (though historically interesting) back-door means of getting pub Possibly, if such groups were made legitimate, recognized,
lication for a mass of writing that might be better lost. and given newspaperlike broadsheet journals circulating to a
86 IN V ISIB L E COLLEGES IN V ISIB L E COLLEGES 87
few hundred individuals, this would spoil them, make them mon to organize research, especially big machine work, around
objects of envy or of high-handed administration and formality. quite a large team of men comprising a few leaders in various
Elite scientific newspapers or broadsheets of this sort have long specialties and a large number of younger men. Now it be
existed in Japan, a country faced with the special problem comes the custom to publish as just such a team. As an editor
that many of its top scientists spend appreciable periods in of Physical Review Letters plaintively noted on a recent oc
foreign institutes. casion, “The participating physicists are not mentioned, not
The scientific elite have acquired prestige among the public even in a footnote.”
in general and the employers in particular, which has given Surprisingly enough, a detailed examination of the incidence
them a certain affluence and enabled them to commute. It of collaborative work in science shows that this is a phenom
incidentally replaces the kudos they have lost since the de enon which has been increasing steadily and ever more rapidly
basement of the coinage of scientific publication. Despite a since the beginning of the century (Fig. 19). It is hard to
tendency to place summer schools in pleasant resort areas find any recent acceleration of the curves that would corre
whenever possible and to make institute housing a good place spond to the coming of the big machine and indicate this as
to bring one’s family, there is a further need. There is a further a recognizable contributing cause.
need to recognize that although a place such as Brookhaven Data from Chemical Abstracts show that in 1900 more
was once where one went to work with big machines and than 80 percent of all papers had a single author, and almost
certain other facilities, it has come nowadays to play an in all the rest were pairs, the greater number being those signed
creasingly important role as a station on the commuting circuit by a professor and his graduate student, though a few are
of several invisible colleges. People come to work with other of the type Pierre and Marie Curie, Cockcroft and Walton,
people, who have come to work with yet other people, who Sherlock Holmes and Dr. Watson.^i since that time the pro
happen to be there. We need many more such facilities in portion of multi-author papers has accelerated steadily and
various fields and in various countries. It might, for example,
“ S. A. Goudsmit, Physical Review Letters, 8 (March 15, 1962), 229.
be wise for the United States government to subsidize the Another good example of a quite different sort of collaboration is the
erection of “Fulbright residential buildings” in London, Cam appearance of the world’s greatest pseudonymous mathematician, Nico
las Bourbaki. This Frenchman with a Greek name, author of an inter
bridge and Oxford, Copenhagen, Geneva, Paris, Delhi, and nationally famous collection of treatises on modem higher mathematics,
wherever else United States scientists habitually commute in is actually a group of 10 to 20 mathematicians, most of them French, all
quantity. of them highly eminent in their fields, none of them identified by name
as part of the polycephalic Bourbaki. See Paul R. Halmos, “Nicolas
So much for the elite, what of the masses? Mention of the Bourbaki,” Scientific American, 196 (May, 1957), 88-99.
big machines is immediately reminiscent of one way in which “ Results of an unpublished investigation by L. Badash, Yale Univer
sity.
the formation of elites is producing a problem in the organiza “ L. Kowarski, “Team work and individual work in research,” CERN
tion of the rest of the scientific population. It has become com Courier, 2 ( May, 1962), 4-7.
IN V ISIB L E COLLEGES 89
powerfully, and it is now so large that if it continues at the
present rate, by 1980 the single-author paper will be e x tin c t .^ ^
It is even more impressive that three-author papers are ac
celerating more rapidly than two-author, four-author more
rapidly than three-author papers, and so on. At present only
about one paper in four has a multiplicity of three or more
authors, but, if the trend holds, more than half of aU papers
will be in this category by 1980 and we shall move steadily
toward an infinity of authors per paper. It is one of the most
violent transitions that can be measured in recent trends of
scientific manpower and literature.
One way of understanding this movement toward mass col
laboration is to see it as a natural extension of the growth
created by the constant shift of the Pareto distribution of
scientific productivities. There is a continuous movement to
ward an increase in the productivity of the most prolific
authors and an increase in the numbers of those minimally
prolific. As we approach a limit in both directions, it is
clear that something has to give. The most prolific people
increase their productivities by being the group leaders of
teams that can accomplish more than they could singly. The
minimal group are in short supply and we can hardly afFord
**Cf. data for Mathematical Reviews and three United States mathe
matics journals ( percent papers having joint authors):
Math. Revs. Three U.S. journals
1920 2.2
1930 4.1
AS A FUNCTION OF DATE 1940 5.8 18.2
1950 6.5 18.2
Data from Chemical Abstracts, 1910-60, are here presented showing the 1960 10.8 12.7
percentages of papers having a single author, those with two, three, and From a letter by W. R. Utz, American Mathematical Society Notices, 9
four or more. It seems evident that there has been a steadily accelerating (1962), 196-97.
change since the beginning of the century.
90 IN V ISIB L E COLLEGES
IN V ISIB L E COLLEGES 91
to let them grow until they reach that ripeness of producing
be effected by the publication of papers. We tend now to com
significant papers on their own. By the creation of a class
municate person to person instead of paper to paper. In the
of fractional authors—that is, scientists who produce one nth
most active areas we diffuse knowledge through collabora
part of a scientific paper—a much larger number of the minimal
tion. Through select groups we seek prestige and the recogni
group is kept at the lower end of the distribution. One expects
tion of ourselves by our peers as approved and worthy col
that as these individuals grow they will evolve into unit authors
laborating colleagues. We publish for the small group, forcing
or better, but in the meantime the body of research workers
the pace as fast as it will go in a process that will force it
is increased to meet demand. It is to some extent accidental
harder yet. Only secondarily, with the inertia bom of tradi
that wartime organization and the advent of the big machine
tion, do we publish for the world at large.
have occasioned the introduction of fractionality, without
All this makes for considerable change in the motivation
which we should have a severe manpower shortage.
of the scientist; it alters his emotional attitude toward his
A more optimistic viewpoint to take is that the emergence
work and his fellow scientists. It has made the scientific paper,
of this class of sorcerer s apprentices partly solves the problem
in many ways, an art that is dead or dying. More than this,
of organizing the lower-level scientists so that they can be
the invisible colleges have a built-in automatic feedback mech
directly related to the research life of the elite. This is nothing
anism that works to increase their strength and power within
but a logical extension of that old familiar principle, the great
science and in relation to social and political forces. Worse,
professor with his entourage of graduate students, the sort
the feedback is such that we stand in danger of losing strength
of thing for which Rutherford or Liebig are well known. The
and efficiency in fields and countries where the commuting
great difference here is that the apex of the triangle is not a
circuit has not yet developed. In short, now that we have
single beloved individual but an invisible college; its locale
achieved a reasonably complete theory of scientific manpower
is not a dusty attic of a teaching laboratory but a mobile
and literature, we must look to the social and political future.
commuting circle of rather expensive institutions. R. E. Weston
et ah have suggested that one might name such teams as the
Dubna Reds and the Harvard M.I.T. Yankees, and give each
player a rating.^®
Because of this, one of the great consequences of the transi
tion from Little Science to Big Science has been that after
three centuries the role of the scientific paper has drastically
changed. In many ways the modem ease of transportation and
the affluence of the elite scientist have replaced what used to
“ Letter in Physics Today, 15 (June, 1962), 79-80.
P O LIT IC A L STRATEGY 93
productive ones, we derive the frightening costly principle
that research expenditure increases as the fourth power of the
number of good scientists. It has already been estimated that
(i) the United States may possess enough talent to multiply the
population of distinguished scientists by a factor of five. Let
POLITICAL STRATEGY us be conservative and envisage a future in which it is only
FOR tripled; we could reach this point quite some time before the
year 2000. By then, according to the principle just derived, our
BIG S C I E N T I S T S expenditure would have multiplied by a factor of 81, and
would thus be more than double our entire Gross National
Product.
IN OUR ANALYSIS of the growth of science we have reached a It seems incontrovertible that such an increase in the cost
basic understanding of normal exponential increase and dis of science has been taking place. National research and de
tribution of talent and productivity. Now let us turn our at velopment expenditures were about three billion dollars in
tention to the abnormal—that is, to those things that do not 1950 and thirteen billion dollars in 1960—more than a doubling
follow the pattern. Without doubt, the most abnormal thing every five years. The 15 percent annual increase must be
in this age of Big Science is money. The finances of science matched against a rise in the Gross National Product of only
seem highly irregular and, since they dominate most of the 3% percent a year. At the present rate, science will be 10 per
social and political implications, our analysis must start here. cent of the Gross National Product as early as 1973. It is al
If the costliness of science were distributed in the same way ready in the region 2 to 3 percent, depending on definition.
as its productivity or excellence, there would be no problem. Let us be optimistic and suppose that growth of the Gross
If the per capita cost of supporting scientists were constant, National Product will continue, with no manpower shortage
we should only spend in proportion to their number, so that to impede the increase in the number of qualified scientists,
the money they cost would double every 10 to 15 years. But and return to the question of whether the cost per scientist
in fact our expenditure, measured in constant dollars, doubles must also increase. Data from the federal agencies that now
every 5% years, so that the cost per scientist seems to have support so much research indicate clearly that the cost per
been doubling every 10 years. To put it another way, the cost project has been rising rapidly. The National Institutes of